Simulating dengue fever transmission process based on weather and environmental factors

• Main Author: Goh, Hui Ping.
• Other Authors: Wang Lipo
• Format: Final Year Project
• Language: English
• Published: 2009
• Subjects: DRNTU::Engineering
• Online Access: http://hdl.handle.net/10356/17890
• Institution: Nanyang Technological University

Climatic factors had been believed to have influences in infectious diseases such as dengue and Hand-Foot-Mouth Disease (HFMD) incidence. Relationships between climatic factors and infectious diseases were investigated to determine disease variations and also assessing the possibility of their use for disease prediction and initiation of emergency response procedures. This model shown in this study would be helpful as it is based on reliable, easily obtained and low cost weather data. Disease incidence and weather variables are collected over the study period of 2001 – 2007. Correlation analysis was performed to quantify relationship between climatic factors and number of cases. Factors with high correlation were then identified to be input into the regression model used for prediction. Support-Vector Machine (SVM) is also used to establish the prediction model of the influence of climate variables with disease incidence. The results were then compared with results obtained from the regression model. Weekly maximum and mean temperature, maximum and mean humidity were positively related to dengue while rainfall was inversely correlated. On the other hand, temperature was mostly inversely correlated with HFMD while humidity, rainfall and wind stress were positively related. Both linear regression and SVM predict dengue incidence well but not HFMD. The prediction precision of SVM is superior to that of linear regression for dengue and HFMD model 2, while linear regression proved to be a better predictor for HFMD model 1. Weather variables have different impact on the transmission of dengue and HFMD. The early warning model, based on the factors used in model 1 and 2 in both linear regression and SVM, was developed to predict disease incidence for up to 8 weeks in advance. This model may not be reliable as a sole predictor for disease in Singapore as there may have other factors affecting the disease incidences. However, it is definitely useful to serve as a reference for future warning system development.